Solute-solute interactions in intermetallic compounds

Transcription

1 Solute-solute interactions in intermetallic compounds Debashis Banerjee, Ryan Murray, and Gary S. Collins Department of Physics and Astronomy, Washington State University, Pullman, WA 99164, USA Received: Keywords: quadrupole interactions, PAC, solute atoms, intermetallics, defect interactions Abstract Two types of solute-solute interactions are investigated in this work. Quadrupole interactions caused by nearby Ag-solute atoms were measured at nuclei of 111 In/Cd solute probe atoms in the binary compound 2 using the method of perturbed angular correlation of gamma rays (PAC). Locations of In-probes and Ag-solutes on both - and -sublattices were identified by comparing site fractions in -poor and -rich 2 (Ag) samples. Interaction enthalpies between solute-atom pairs were determined from temperature dependences of observed site fractions. Repulsive interactions were observed for close-neighbor complexes In//+Ag// and In//+Ag// pairs, whereas a slightly attractive interaction was observed for In//+Ag//. Interaction enthalpies were all in the range ±0.15 ev. Temperature dependences of site fractions of In-probes on locally defect-free - and -sites yields a transfer enthalpy that was found to be ev in a previous study of undoped 2. The corresponding values in 2 (Ag) samples are much smaller. This is attributed to competition of In- and Ag-solutes to occupy sites of the same sublattice. While the difference in site-enthalpies of In-solutes on - and -sites is temperature independent, it is proposed that the transfer of Ag-solutes from - to -sites leads to a large temperature dependence of degeneracies of levels available to In-solutes, resulting in an effective transfer enthalpy that is much smaller than the difference in site-enthalpies. Permanent address: Accelerator Chemistry Section, RCD(BARC), Variable Energy Cyclotron Centre, Kolkata , India 1

2 Introduction There is considerable interest in the site preference of solute atoms in compounds. A previous study was carried out of indium solute atoms in 2 using perturbed angular correlations of gamma rays, or PAC, as a function of the detailed composition and temperature of the compound [1]. 2 has the cubic MgCu 2 Laves structure and is highly ordered. The -sites have cubic point symmetry and the -sites have axial symmetry. This allows one to readily distinguish occupations of the two sites by measuring nuclear quadrupole interactions. Measurements revealed that the site fraction of In probe atoms on -sites, In, increases in -poorer samples, consistent with the heuristic rule that impurity atoms tend to occupy the sublattice of an element in which there is a deficiency [1]. In addition, In-probes in thermal equilibrium were observed to transfer from - to -sites with increasing temperature, in effect forming a two-level quantum system in which the enthalpy of a probe on the -site was apparently 0.343(3) ev higher than on the -site [1]. A second subject of interest in the theory of solutions is the study of interactions between pairs of solute atoms. Interactions between solute atoms in pure metals were studied extensively in the 1980 s by Krystof Królas and coworkers using PAC [2, 3, 4, 5 ] and by others using Mössbauer effect [6, 7]. Metals have a unique lattice location and interactions were observed, for example in Ag-metal, between a PAC probe such as 111 In/Cd and close neighbor solutes such as Pd, In, or Sn. The 111 In PAC probe was highly dilute (mole fraction ~10-11 ). A favorable solute mole fraction used was ~1 at.%, so that in the absence of interaction between the probe and solute, there would be a ~10% site fraction for probes having a single solute neighbor, which can be readily measured. If the probe-solute interaction is attractive, the probe-solute site fraction will be large at low temperature and decrease with increasing temperature; if it is repulsive, the fraction will be small at low temperature and increase with temperature. Interaction enthalpies were determined from temperature dependences of the ratio of the site fractions of probes having one or zero neighboring solutes: 2

3 f f zcexp( S / k )exp( Q / k ), (1) 1 B BT 0 in which z is the number of near neighbors, c is the solute mole fraction, and S and Q are the vibrational entropy and enthalpy of interaction. (Q is negative in the case of an attractive interaction.) Królas was able to explain observed interaction enthalpies as well as electric-field gradients caused by the neighboring solute atoms using a model of screened coulomb potentials [3]. Effective charges of solutes based on the difference between the nominal valence of a solute and the host atom it replaced proved useful to explain the results. Consider, for example, the interaction between In probes and Pt solutes in Ag [2]. Using nominal valences of +3, 0, and +1 for In, Pt and Ag, Królas obtained effective charges of +2 for In and -1 for Pt relative to the Ag-host atoms replaced, resulting in an attractive screened interaction. Likewise, probe and solute atoms exhibited repulsive interactions when both had higher valences than host atoms. In the present work, the same approach is applied to investigate solute interactions in ordered binary compounds. Unlike in pure metals, an impurity atom will exhibit a preference to occupy sites of one element or the other. The effective charge of an impurity atom will now depend on the sublattice it occupies. This applies to both solute and probe atoms, so that the interaction enthalpy will differ depending on the sites occupied. This is illustrated in the present work, in which the interaction between In and Ag impurities in 2 is variously attractive or repulsive. 2 (Ag) was chosen for study because the site preference of 111 In probes had previously been extensively studied and because it was found that In has an appreciable occupation of both sublattices [1]. The systems studied are strictly quaternary compounds, but the mole fraction of 111 In/Cd probes is so low, ~10-11, that they do not make binary phases with other constituents. An appropriate system is a binary compound like 2 into which ~1 at.% of a solute dissolves onto one or other or both sublattices. Extensive unsuccessful efforts to identify this kind of ternary phase were made before 2 (Ag) was found. This work is ongoing. Studies in 2 using other solutes are planned. A goal of such studies will be to elucidate rules governing interaction enthalpies between solutes on the various sublattices of a compound and to predict whether 3

4 interactions are attractive or repulsive and how they depend on the nature of the host and solute elements. 2 has the cubic Laves Cu 2 Mg structure, with - and -sites in the perfect crystal having 43m (cubic) and 3 m (axial) point symmetries, respectively [8]. The cubic site has zero electric field gradient (EFG) and therefore a quadrupole interaction frequency of zero. The axial site has a nonzero quadrupole interaction frequency. In the present work, samples of 2 were doped with 111 In activity and ~1-2 at.% of Ag. Quadrupole interactions were observed for In-probes forming complexes with neighboring solute atoms. While PAC measurements of quadrupole interactions are made for the 247 kev PAC level of the daughter 111 Cd nuclide, it should be noted that it is the interaction enthalpy between the 111 In parent probe and solute atom that is measured. Temperature dependences of the ratio of site fractions of probes with and without a solute neighbor were plotted and interaction enthalpies were determined by fitting to eq. 1. This enthalpy is the change in enthalpy when the solute atom is close to, or far away from, the probe. Experimental Samples of 2 were doped with 111 In activity (mole fraction ~ ) and ~1-2 at.% of Ag by melting together all constituents under argon in a small arc-furnace. Metal purities were 3N for and 4N for, and the 111 In activity was carrierfree. For each solute concentration, two samples were made that were modestly -poorer or -poorer than the stoichiometric composition. The difference was expected to promote occupation of one sublattice or the other by the solute atoms and probes according to the heuristic rule. Sample compositions Ag 1.0 and Ag 1.0 were determined from masses of the elements measured prior to melting, and taking into account small mass losses during melting. 111 In decays to the second excited state of 111 Cd by electron-capture with a mean life of 4.0 days. 111 Cd subsequently decays to the ground state with emission of 173 and 247 kev gamma rays, the 247 kev PAC level having a mean life of 120 4

5 ns. PAC measurements were made using a four-detector PAC spectrometer employing 1.5 x 1.0 inch BaF 2 scintillators. Coincidences detected between the 173 and 247 kev gamma rays give lifetimes of the PAC level and were histogrammed over typical measurement times of one day. For each measurement, four time-coincidence spectra were accumulated simultaneously, two each at detector angles of 180 o and 90 o relative to the sample. After fitting and subtracting accidental backgrounds, spectra were geometrically averaged and combined to yield the experimental PAC spectrum. The PAC spectrum was fitted with a superposition of quadrupole perturbation functions, one for each distinct site occupied by probe atoms: G ( t) s n n t sn cos( nt)exp. (2) The perturbation function has four terms, with the three non-zero frequency components arising from hyperfine splitting of the spin I= 5/2 PAC level in an EFG. They have the properties that and The frequencies depend on the quadrupole interaction strength and the asymmetry Vxx Vyy parameter of the traceless EFG tensor,, in which V zz is its principal V zz component, so that 0 1. The fundamental observed frequency, 1, is a function of the magnitude of the quadrupole interaction and asymmetry parameter. For the special case of axial symmetry, 0, : 1: 2 : 3, : and 1 3 eqv zz 10 h, in which Q is the quadrupole moment of the PAC level and h is Planck s constant. For each signal, the amplitudes sum to unity, 1. The amplitude of each fitted perturbation function (eq. 2) is the site fraction for that signal. describes inhomogeneous broadening of the signals due to weak EFG disturbances from distant defects. 3 s n n 0 1 is used below to label the observed signals. For additional information about PAC spectroscopy and methodology, see [9]. 5

6 Results Six signals were detected in Ag-doped 2 and found consistently in many measurements. They are listed in Table 1 and identified below by average fundamental quadrupole interaction frequencies 1. The 44, 0 and 9 Mrad/s signals were observed in the earlier study of site preferences in undoped 2 [1], where they were identified instead by their frequencies measured at room temperature, 49, 0 and 8 Mrad/s, respectively. The dominant 0 and 44 Mrad/s signals were attributed to isolated In-probes on - and -sites, consistent with the point symmetries of the sites [1]. The 9 Mrad/s signal was only observed in -poor samples and was reasonably attributed to an In-probe on -site with a neighboring antisite defect. The 54, 63 and 76 Mrad/s signals were not observed in the previous study and are attributed to close neighbor complexes of In-probes with Ag-solutes. The table gives attributions of the signals to complexes on the basis of composition dependences presented in Fig. 2 below. so listed in the table are the types of samples in which the signals were observed and measured interaction enthalpies. Table 1. Quadrupole interaction parameters 1 (average value over all measurements) and. Column 3 lists samples in which the signal is observed. Column 4 gives attributions discussed in the text. Column 5 gives the interaction enthalpy Q between the two impurity atoms based on fits to eq. 1 (with a negative sign indicating an attractive interaction) and have uncertainties of order 0.04 ev. 1 (Mrad/s) Samples Attribution Interaction Enthalpy 44 ~0 Pure and doped 0 ~0 Pure and doped 9 ~0 Pure and doped, -poor 54 ~0 Ag-doped, -poor Ag-doped, -rich 76 ~0 Ag-doped, -rich and poor In - In - In ev (pure); ev (doped) In Ag ev In Ag ev In Ag ev 6

7 Representative spectra are shown in Fig. 1, with the time-domain PAC spectrum on the left. (The data at apparent negative times in the figure are independent data that mirror the positive delayed-coincidence time spectrum. They are obtained by gating times of arrival of the 173 and 247 kev gammas in reverse order. Such double-sided spectra make the shape of the spectrum near time zero much clearer.) Frequency transforms are shown at right, with tridents indicating the three frequency components of the 44 and 54 Mrad/s signals. The results are now presented and discussed in several paragraphs Ag C 0Amplitude (a.u.) G 2 (t) C 40Time (ns) C C 054 Frequency (Mrad/s) 030Figure 1. Representative PAC spectra for a (31)(67)Ag(2) sample measured at 300 and 700 o C, with time-domain spectra on the left and frequency spectra on the right. Two significant signals are identified at right by tridents that indicate the three frequency components. In addition, a vertical offset representing a zero-frequency signal is seen to be greater in the spectrum at 300 o C. A. Boundary compositions of the 2 phase. 2 appears on binary phase diagrams as a line compound, having a phase field whose width has not been measured and is probably less than about 1 at.%. The more -rich boundary composition is likely to be close to the 33.3% stoichiometric composition since no signal was observed in the previous work [1] that could be attributed to a antisite defect. This is not unexpected since the atomic volume of is much greater than of, leading to a large strain 7

8 interaction and a high expected defect formation enthalpy for the antisite defect. The -poorer boundary composition is likely to have a fixed value below the stoichiometric composition in the range at.%. B. Attributions of signals to specific probe-solute complexes based on the composition dependence of site fractions. Fig. 2 shows fitted site fractions of the observed signals measured at 400 o C plotted versus nominal average compositions of the alloys. It can be seen that some sample compositions probably lie outside the likely boundary compositions discussed above in (A). This should lead to small volume fractions of secondary phases that would produce small site fractions of additional signals that go undetected. However, no such signals were observed in this or in the previous study [1]. There follow rationales for attributions of the 54, 9, 63 and 76 Mrad/s signals. 3Site Fractions (%) content (%) content (%) at.% Ag 0f at.% Ag f44 f76 f0 f at.% Ag f at.% Ag f9 f54 f76 f9 f0 0f63 f93f93site Fractions (%) Figure 2. Site fractions of signals observed in -richer and -poor samples having either 1.0 or 2.0 at.% of Ag-solute. Signals are identified by quadrupole interaction frequencies given in Table 1; thus, "f9" is the site fraction for the 9 Mrad/s signal. The figure at left shows the composition dependence of the 44, 0 and 9 Mrad/s signals. At right are shown dependences for the three 8

9 signals associated with Ag-solutes as well as for the 9 Mrad/s signal previously attributed [1] to the complex In. The 54 and 9 Mrad/s defects with 0 are attributed to defects on the -sublattice neighboring In probes. The 54 Mrad/s defect was observed only in the -poorer samples. Similarly, the site fraction of the 9 Mrad/s signal is significantly smaller in -richer samples. Both signals have axial symmetry (zero asymmetry parameter), consistent with a single defect near an In probe. The 9 Mrad/s signal was previously observed in undoped 2 and identified with a In complex [1]. Appearing only in Ag-doped samples, the 54 Mrad/s signal is similarly identified with a In Ag complex. Magnitudes of the frequencies (or EFGs) are also consistent with the small effective charge of an defect ( and have the same nominal valences) and a larger anticipated effective charge for an Ag defect (Ag and have valences differing by 2). The 63 Mrad/s, =0.3 defect is attributed to a close neighbor pair. In Ag The 63 Mrad/s site fraction disappears at the -poorer phase boundary in both Ag-doped samples. Site fractions for both the 44 and 63 Mrad/s become smaller at the -poorer boundary composition, which is consistent with this attribution. The nonaxial EFG of the complex indicates that it is a superposition of two noncollinear EFGs. The lattice EFG at the undecorated site is along the <111> direction in the cubic unit cell while the polar angle from an site to a neighboring -site is 25.7 o, leading to a nonaxial EFG. It can be shown that all near-neighbor pairs In In Ag will have the same EFG. 9

10 The 76 Mrad/s signal with 0 is attributed to a close neighbor In Ag pair. The site fraction of the 76 Mrad/s complex remains roughly the same at the richer and -poorer boundary compositions. Having 0 suggests that the complex involves In probes. At the same time, the site fraction for undecorated In is smaller by a factor of three at the -richer boundary. Only the complex In Ag explains these observations satisfactorily: although the concentration of concentration of In Ag decreases as the composition becomes more -rich, the increases. A fourth anticipated but unobserved complex is In Ag. It would be expected to have site fractions that were similar in -rich and -poor samples, in the same way as for the In Ag complex with 76 Mrad/s frequency, but no such signal was identified. C. Competition of host and solute atoms to occupy sites. In pure 2, measurements were made on five samples of slightly different composition. An average activation enthalpy (3) ev was observed for the transfer of In-probes from -sites to -sites [1]. Labeling the sites with their quadrupole interaction frequencies 0 and 44 Mrad/s, respectively, the temperature dependence of the site-fraction ratio is given by f0 / f44 f / f ~ exp( Q / k T ), (3) B in which Q is an enthalpy of transfer. Figure 3 shows corresponding Arrhenius plots of the ratio of site-fractions of In-probes on - and -sites for four samples having 1 or 2 at.% of Ag-solute and that were -poor (~31.3 at.%) or -rich (~33.3 at.%). 10

11 (2.0 at.%ag)111213( (80q = 0.28 (3) ev Q = 0.22 (3) ev f 0 /f 44 0.T (1.0 at.%ag) Q = 0.04 (8) ev 8K0700)31.5 (1.0 at.%ag).ev1evt)f9/f0 K031.5 (1.0 at.%ag) 31.0 (2.0 at.%ag) (2.0 at.%ag) 1Q = 0.10 (0.14) ev 141/k B T (ev -1 ) 15141/k B T (ev -1 ) 15Figure 3. (Left) Arrhenius plots of the ratio of site fractions of In-probes on and sites for samples with 1 or 2 at.% Ag and which were more -poor (top) or -rich (bottom). The transfer enthalpies are of order 0.25 ev (top) and 0.07 ev (bottom), much smaller than the value ev observed for undoped 2 [1]. (Right) Arrhenius plots of the ratio of site fractions of the In complex and undecorated In probe for samples with 1 and 2 at.% Ag. It can be seen in Fig. 3 (left) that the corresponding activation enthalpy is ~0.25 ev for Ag-doped, -poor samples and ~0.07 ev for Ag-doped, -rich samples. The measured transfer enthalpies are compared in Figure ev 0.25 ev 0.07 ev Undoped Ag-doped, -poor Ag-doped, -rich Figure 4. Schematic of measured transfer enthalpies in three types of 2 samples. 11

12 Naïvely, the transfer enthalpy should be equal to the difference in enthalpies of the solute at the two sites. But site-enthalpies of isolated solutes like these Inatoms depend solely on local atomic environments within 1-2 atomic shells, and therefore should be essentially independent of the mole fractions of solutes or of other defects such as antisite atoms or lattice vacancies. The large observed changes demonstrate that the transfer enthalpy is not simply equal to the difference in site-enthalpies. To explain the changes in the transfer enthalpies shown in Fig. 4, it is proposed that degeneracies of the and -levels are strongly temperature-dependent, so that the activation enthalpies for transfer of In-solutes between the two sublattices shown in Fig. 3 or 4 are not simply equal to the difference in site-enthalpies. Eq. 3 can be is generalized to include degeneracies of sites available to In-solutes on the two sublattices: f f 44 f g ( T ) exp( Q / kbt) exp( Q' / kb ), (4) f g ( T ) 0 T in which the g(t) s are temperature-dependent degeneracies of the two sets of levels. The observed reductions in the transfer enthalpy indicate that the ratios g ( T ) / g ( T ) of the degeneracies must increase with temperature. If one assumes that g (T ) is constant, then (T ) must decrease with increasing g temperature in the Ag-doped samples. This is illustrated in Table 2, which lists the measured transfer enthalpies, reductions in transfer enthalpies below the value in undoped 2 (which is possibly equal or close to the difference in enthalpies of In on the two sites), and ratios of degeneracy factors calculated for 700K and 950K, the approximate minimum and maximum temperatures of measurement. Table 2. Transfer enthalpies Q from Fig. 3, summarized in Fig. 4, reductions in transfer enthalpies below the value for undoped 2, and ratios of degeneracy factors for - and -sites at 950K and 700K. Sample Q (ev) Q (ev) g g (950K) / g (950K) / g (700K (700K) Undoped [ref. 1] (0) (1) Ag-doped, -poor

13 Ag-doped, -rich The table shows that, relative to the undoped samples, the ratio of degeneracies has changed by factors of 1.5 and 3.2 between 750 and 900 K in the two Agdoped samples. This is consistent with the change in compositions of the samples; the effect of increasing the average -composition is to create antisite atoms, thereby "filling in" virtual vacancies on the -sublattice and impeding transfer of Ag-atoms to the -sublattice. The same effect would occur in the undoped sample, but the mole fraction of In-solutes is so minuscule that their transfer is not impeded. For Ag-doped samples, Ag-solutes switch from -sites to -sites with increasing temperature, so that fewer -sites are available for In-solutes. This implies that Ag- and In-solutes have the same ordering of site-energies. If the ordering were opposite, with Ag-solutes transferring from -sites at low temperature to -sites at high temperature, then the effective transfer enthalpy of In-solutes would have increased relative to undoped 2. The fact that the reduced transfer enthalpies in the Ag-doped samples were approximately the same for 1 and 2% mole fractions of Ag indicates that 1 at.% Ag is sufficient to cause the reduction in the degeneracy ratio. Fig. 3 (right) shows an Arrhenius plot of the ratio of site fractions of the In complex and undecorated In probe for -poor samples having 1 and 2 at.% Ag and ~31.3 at.%. The interaction enthalpies of ev (2% Ag) and ev (1% Ag) can be compared with -0.16(4) ev (0% Ag) observed for an undoped 2 sample with composition 33.3(1) at.% [sample B in ref. 1]. The change in sign and increase in transfer enthalpy in going from 0 to 1 to 2 at.% Ag is considered to result from cross-interaction of Ag-solutes and host atoms. It was already shown above that Ag atoms transfer from -sites to sites with increasing temperature. This transfer in turn may displace -atoms from their sublattice to the -sublattice. If the increase in concentration of with temperature is sufficiently large, it may change the sign of the effective interaction enthalpy of the complex In from ev (attractive) to to 0.24 ev (an apparently repulsive interaction), solely due to an increase in 13

14 1()the mole fraction of antisite defects. The difference in the interaction enthalpies gives the effective activation enthalpy for the increase in the concentration of, or about ev (1 at.% Ag) to ev (2% Ag). This scenario appears to explain the observations. As an alternative explanation, the ~9 Mrad/s signal might be composite and include signals having similar low frequencies from both In and the Ag In complex not yet accounted for. But it is unlikely that the frequency would be so much lower than the 44 Mrad/s frequency of an isolated In probe atom. D. Probe-solute interaction enthalpies. Interaction enthalpies for the three In-Ag pairs were obtained by fitting ratios of site-fractions according to eq. 1. Consider first the interaction between members of the In Ag pair of solute atoms, with 54 Mrad/s frequency. The activation enthalpy to form the pair can be determined from the temperature dependence of the ratio f( In Ag )/f( ), or f 54 /f 0, using Eq. 1. This is shown in Figure 5 for -poor samples with 1 or 2 at.% Ag. The average activation enthalpy was ev, indicating a repulsive interaction. One notes that the site fraction is approximately twice larger for c= 2% of solute than 1%, as expected from Eq. 1. TK In (1.0 at.%ag) 31.0 (2.0 at.%ag) Q = 0.14 (8) ev f /f 54 0 Q = 0.12 (6) ev /k B T (ev -1 )

15 1(Figure 5. Arrhenius plot of the ratio of site fractions of complex In Ag and isolated probe In. Interaction enthalpies Q were obtained by fitting with Eq. 1 and are positive, indicating a repulsive interaction. Consider as a second example the Arrhenius plot of the ratio of site fractions for the complex In Ag (76 Mrad/s) and isolated probe (0 Mrad/s) shown Q = 0.20 (6) ev in Fig. 6. The interaction enthalpies are all of order ev, indicating a repulsive interaction..1in0tk )31.5 (1.0 at.%ag) (2.0 at.%ag)1q = 0.14 (6) ev f 76 /f Q = 0.12 (21) ev (1.0 at.%ag) 33.0 (2.0 at.%ag) 0.1Q = 0.19 (14) ev 1/k B T (ev -1 ) Figure 6. Arrhenius plot of the ratio of site fractions of complex In Ag and isolated probe In. Interaction enthalpies Q were obtained by fitting with Eq. 1 and are positive, indicating a repulsive interaction. Table 1 lists all the interaction enthalpies, including for the third Ag-In complex, In Ag, which has an attractive interaction, and for the intrinsic -antisite complex, In, discussed in Section C above. The magnitudes of all interaction enthalpies are small, in the range ±0.15 ev, typical of what has been observed or calculated for PAC and Mössbauer studies in pure metals [ 2-7]. 15

16 Summary and Conclusions. Experiments were carried out to determine solute-solute interactions in the intermetallic compound 2 using PAC spectroscopy. One solute was the 111 In PAC probe, present at a mole fraction of The other solute was 1-2 at.% of Ag. Three different close-atom pairs of Ag and In solutes were detected by characteristic EFGs. Identification of the pairs was made by examining the observed symmetry of the EFGs and how their site-fractions varied with composition. In and Ag solutes were found to populate both - and sublattices. Interaction enthalpies of the pairs were attractive or repulsive, and in the range ±0.15 ev, typical of what has been seen in the pure metals. In a previous study of undoped 2, a transfer enthalpy for In-solutes of ev was observed between sites on the - and -sublattices. In the present study, Ag-doping was found to lead to a large reduction in the effective transfer enthalpy for In-solutes between - and -sublattices. It is proposed that this reduction is a consequence of large changes with temperature of the degeneracies of levels available for In-solutes on - or -sublattices, with changes in the degeneracies caused by thermally-activated transfer of Ag-solutes also between - and -sublattices. The observed trends are consistent with Agsolutes also tending to transfer from the -sublattice at low temperature to the -sublattice at high temperature. Through measurements such as this, it is possible to determine the direction of flow of a solute in a compound with increasing temperature. Acknowledgement. This work was supported in part by the National Science Foundation under grant DMR (MMN Program). Krystal Kasal (MS, 2015) helped carry out early experiments to identify appropriate ternary systems for further study. 16

17 1 Matthew O. Zacate and Gary S. Collins, Physical Review B69, (1-9) (2004). 2 A.Z. Hrynkiewicz and K. Krolas, Physical Review B28, (1983). 3 M. Sternik and K. Krolas, Physical Review B40, ( K. Krolas, W. Bolse and L. Ziegeler, Hyperfine Interactions 35, 635 (1987). 5 M. Sternik and K. Królas, Acta Physica Polonica A82, (1992). 6 T.E. Cranshaw, J. Phys. F: Met. Phys. 17, (1987). 7 J. Chojcan, Journal of loys and Compounds 264, (1998). 8 P. Villars and L.D. Calvert, Pearson s Handbook of Crystallographic Data for Intermetallic Compounds, 2 nd ed. (ASM International, Materials Park, Ohio, 1991). 9 G. Schatz and A. Weidinger, Nuclear Condensed-Matter Physics, (John Wiley, New York, 1996). 17